Online Fake Comments Detecting Model Based on Feature Analysis

Abstract

Contraposing to the existing problems in fake comments detecting, an online fake comments detecting model is proposed with the dynamic information contained in historical behaviors of user. The scheme makes adopts sequence analysis model to mine the dynamic features of users from dynamic information. By the idea of semi supervised learning, two kinds of features are taken as independent views, which are used to establish classifier and to choose unlabeled samples with high confidence. Then these selected samples are used to update training model and improve the effects of classifier. Finally through the analysis of abnormal behavior of product reviews, we filter fake comments of commodities to provide more accurate research object for fake comments analysis.